Target association method and device, computer device and storage medium

Abstract

The application relates to the technical field of target detection, in particular to a target association method and device, computer equipment and a storage medium. The method comprises the following steps: extracting a first target object from first sensing data collected by a first sensor, and extracting a second target object from second sensing data collected by a second sensor; adjusting a reference association door according to a first distance between the first target object and the first sensor and a second distance between the second target object and the second sensor, to obtain a dynamic association door; and determining whether the first target object and the second target object belong to the same object based on the dynamic association door. The application can improve the accuracy of target association.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more particularly to the fields of autonomous driving, intelligent transportation and target detection technology, specifically to a target association method, apparatus, computer device and storage medium. Background Technology

[0002] Autonomous vehicles encompass modules such as perception, decision-making, and control. Among these, environmental perception involves various sensors, such as cameras, millimeter-wave radar, and lidar, which form the basis for target detection tasks in autonomous driving scenarios.

[0003] When tracking targets detected by various sensors, data association algorithms are typically used. For the same target, the measurements generated by different sensors will inevitably have certain similar characteristics due to their shared physical origin. However, due to noise interference and the instability of the sensors themselves, the characteristics of these measurements are not entirely identical. The purpose of data association is to utilize these similar characteristics to determine whether these measurement data originate from the same target. One of the logical principles of data association is that each measurement observation can only be associated with one other measurement, and cannot be associated with two or more other measurements simultaneously.

[0004] However, the detection accuracy of sensors is affected by the detection environment and can fluctuate significantly, making it difficult for the above algorithm to accurately correlate the targets tracked by two sensors at the same time. Summary of the Invention

[0005] Therefore, it is necessary to provide a target association method, apparatus, computer device, and storage medium that can improve the accuracy of target association in order to address the above-mentioned technical problems.

[0006] Firstly, this application provides a target association method, which includes:

[0007] Extract a first target object from the first sensing data collected by the first sensor, and extract a second target object from the second sensing data collected by the second sensor;

[0008] Based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, the baseline association gate is adjusted to obtain the dynamic association gate;

[0009] Based on dynamic association gates, it is determined whether the first target object and the second target object belong to the same object.

[0010] In one embodiment, a baseline correlation gate is adjusted based on a first distance between a first target object and a first sensor, and a second distance between a second target object and a second sensor, to obtain a dynamic correlation gate, including:

[0011] Select at least one of the position error correlation gate, the size error correlation gate, and the velocity error correlation gate as the reference correlation gate;

[0012] The observation distance is determined based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor;

[0013] Based on the observation distance and the baseline distance corresponding to the baseline correlation gate, the baseline correlation gate is adjusted to obtain the dynamic correlation gate.

[0014] In one embodiment, the baseline correlation gate is adjusted based on the observation distance and the baseline distance corresponding to the baseline correlation gate to obtain a dynamic correlation gate, including:

[0015] The ratio between the observed distance and the baseline distance is used as an adjustment factor;

[0016] Based on the adjustment factor, the baseline correlation gate is scaled to obtain the dynamic correlation gate.

[0017] In one embodiment, determining whether the first target and the second target object belong to the same object based on a dynamic association gate includes:

[0018] Based on dynamic association gates, the degree of target association between the first target object and the second target object is determined;

[0019] Based on whether the target relevance meets the preset relevance conditions, determine whether the first target object and the second target object belong to the same object.

[0020] In one embodiment, determining whether a first target object and a second target object belong to the same object based on whether the target relevance meets a preset relevance condition includes:

[0021] If the correlation degree meets the correlation conditions corresponding to the dynamic correlation threshold, then the target correlation degree between the first target object and the second target object is determined.

[0022] Based on whether the target relevance meets the preset relevance conditions, determine whether the first target object and the second target object belong to the same object.

[0023] In one embodiment, the method further includes:

[0024] Determine a first reference correlation degree between the first target object and the second reference object; wherein, the second reference object is an object other than the second target object extracted from the second sensing data;

[0025] Determine a second reference correlation degree between the second target object and the first reference object; wherein, the first reference object is an object other than the first target object extracted from the first sensing data;

[0026] If the target correlation degree is less than both the first reference correlation degree and the second reference correlation degree, then the target correlation degree is determined to meet the preset correlation condition.

[0027] In one embodiment, extracting a first target object from first sensing data acquired by a first sensor and extracting a second target object from second sensing data acquired by a second sensor includes:

[0028] Extract a first candidate object from the first sensor data, and extract a second candidate object from the second sensor data;

[0029] Based on the fixed association gate, the first candidate object and the second candidate object are filtered to obtain the first target object and the second target object.

[0030] Secondly, this application also provides a target association device, characterized in that the device comprises:

[0031] The acquisition module is used to extract a first target object from the first sensing data collected by the first sensor, and to extract a second target object from the second sensing data collected by the second sensor.

[0032] The adjustment module is used to adjust the reference association gate according to the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, so as to obtain the dynamic association gate;

[0033] The association module is used to determine whether the first target object and the second target object belong to the same object based on the dynamic association gate.

[0034] Thirdly, this application also provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0035] Extract a first target object from the first sensing data collected by the first sensor, and extract a second target object from the second sensing data collected by the second sensor;

[0036] Based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, the baseline association gate is adjusted to obtain the dynamic association gate;

[0037] Based on dynamic association gates, it is determined whether the first target object and the second target object belong to the same object.

[0038] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0039] Extract a first target object from the first sensing data collected by the first sensor, and extract a second target object from the second sensing data collected by the second sensor;

[0040] Based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, the baseline association gate is adjusted to obtain the dynamic association gate;

[0041] Based on dynamic association gates, it is determined whether the first target object and the second target object belong to the same object.

[0042] The aforementioned target association method, apparatus, computer equipment, and storage medium, by combining a first distance between a first sensor and a first target object detected by the first sensor, and a second distance between a second sensor and a second target object detected by the second sensor, dynamically adjust a baseline association gate to introduce a dynamic association gate corresponding to the distance factor. When the first and second sensors have large detection errors due to distance factors in the environment, this dynamic association gate can reduce the impact of distance factors on target detection and improve the accuracy of target association. Furthermore, preliminary screening using a fixed association gate improves computational efficiency. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the target association method in one embodiment;

[0044] Figure 2 This is a schematic diagram of the process for determining the first target object and the second target object in one embodiment;

[0045] Figure 3 This is a flowchart illustrating the process of adjusting the reference correlation gate based on the observation distance in one embodiment;

[0046] Figure 4 This is a flowchart illustrating the process of determining target relevance in one embodiment;

[0047] Figure 5This is a flowchart illustrating the process of determining whether the target correlation degree meets the preset correlation conditions in one embodiment.

[0048] Figure 6 This is a flowchart illustrating the process of determining the error region in one embodiment;

[0049] Figure 7 This is a schematic diagram of the error region in one embodiment;

[0050] Figure 8 This is a flowchart illustrating the target association method in another embodiment;

[0051] Figure 9 This is a structural block diagram of the target association device in one embodiment;

[0052] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] Autonomous vehicles encompass modules such as perception, decision-making, and control. Among these, environmental perception involves various sensors, such as cameras, millimeter-wave radar, and lidar, which form the basis for target detection tasks in autonomous driving scenarios.

[0055] When tracking targets detected by various sensors, data association algorithms are typically used. For the same target, the measurements generated by different sensors will inevitably have certain similar characteristics due to their shared physical origin. However, due to noise interference and the instability of the sensors themselves, the characteristics of these measurements are not entirely identical. The purpose of data association is to utilize these similar characteristics to determine whether these measurement data originate from the same target. One of the logical principles of data association is that each measurement observation can only be associated with one other measurement, and cannot be associated with two or more other measurements simultaneously.

[0056] However, during the detection process, multiple sensors are often affected by various factors, leading to uncertainties in the fused data. These factors include the distance between the sensor and the object being detected, detection accuracy, transmission medium, computational errors, and noise. If the detection accuracy of the sensors fluctuates significantly, the algorithm may struggle to accurately correlate targets tracked by two sensors at the same time.

[0057] Based on this, this application provides a target association method, applicable to situations where target objects detected by two sensors are associated in a target detection scenario. Optionally, this method can be executed by a computer device, which can be a server or a terminal device, such as an in-vehicle terminal. Specifically, in one embodiment, as... Figure 1 As shown, the method specifically includes the following steps:

[0058] S101, extract the first target object from the first sensing data collected by the first sensor, and extract the second target object from the second sensing data collected by the second sensor.

[0059] In this embodiment, the first sensor and the second sensor are different types of sensors installed on the autonomous vehicle, such as any two of cameras, millimeter-wave radar, and lidar. Optionally, the data acquisition ranges of the first sensor and the second sensor overlap in this embodiment.

[0060] Furthermore, the first and second sensor data are acquired at the same time. In one possible implementation, the first and second sensor data can be data directly obtained from the corresponding sensors to describe the characteristic information of each object; in another possible implementation, the first and second sensor data can also be data obtained by further analyzing the data from the corresponding sensors using a feature extraction algorithm to describe the characteristic information of each object. Specifically, the object characteristic information may include size information, velocity information, position information, and type information parameters.

[0061] It is understandable that, since different sensors use different coordinate systems, it is necessary to transform the data from each sensor into the same coordinate system to achieve spatial alignment of the data. Therefore, the data from the first sensor and the data from the second sensor can be spatially aligned data. In this case, the positions of the first target object and the second target object are in the same coordinate system.

[0062] For example, in this embodiment, the extracted first target object and the second target object both have corresponding object feature information. When associating targets, the object feature information of the first target object and the object feature information of the second target object are used to perform the association operation.

[0063] Furthermore, the number of first target objects and the number of second target objects are both at least one. Specifically, if there are multiple first target objects and multiple second target objects, then this target association method is specifically used to associate multiple targets.

[0064] S102, the reference association gate is adjusted according to the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, to obtain the dynamic association gate.

[0065] In this embodiment, the benchmark correlation gate can be determined based on historical statistical data. It is used to describe the maximum permissible deviation between the first measurement value (e.g., the first target object) and the second measurement value (e.g., the second target object) in a specified dimension. The historical statistical data is used to characterize the deviation between object feature information collected by different sensors for the same target at different historical moments. Furthermore, a benchmark correlation gate can be set based on each feature dimension in the object feature information, such as a correlation gate for the position dimension, a correlation gate for the velocity dimension, a correlation gate for the size dimension, etc.

[0066] Under certain circumstances, the first and second sensors may have significant errors. These errors could be caused by factors such as sensor location (including the distance between the sensor and the target object), detection accuracy, transmission medium, calculation errors, and noise. In this embodiment, to improve the accuracy of target association when sensor errors are large, distance factors are first eliminated from the list of causes of error. That is, based on the first and second distances, the baseline association gate is adjusted to generate a dynamic association gate that can adaptively match distance factors.

[0067] Optionally, in one implementation, all objects in the first sensor are taken as first target objects, and similarly, all objects in the second sensor are taken as second target objects; based on each first target object and each second target object, corresponding object groups are formed, and corresponding dynamic association gates are generated for each object group.

[0068] In another possible implementation, all objects in the first sensor can be initially filtered using corresponding filtering conditions to obtain the first target object. Similarly, all objects in the second sensor can be filtered using corresponding filtering conditions to obtain the second target object. Based on each first target object and each second target object, corresponding object groups are formed, and a corresponding dynamic association gate is generated for each object group. Specifically, the dynamic association gates generated for each of the above object groups are shown in Table 1 below:

[0069] Table 1

[0070]

[0071] In this table, N rows represent N second target objects, and M columns in Table 1 represent M first target objects.

[0072] Specifically, when adjusting the baseline association gate based on the first and second distances to generate a dynamic association gate that can adaptively match distance factors:

[0073] In one possible implementation (Method 1), for any object group, the farther the object group is from the autonomous vehicle (greater than a preset baseline distance range), the lower the recognition accuracy of the first and second sensors for the object group. This may lead to a large error between the first target object (object feature information of the first target object) and the second target object (object feature information of the second target object) detected by the first and second sensors. Given the error in the data source, if a fixed association gate is used to filter the object group, there may be a large error when associating the first and second target objects. For example, when using a fixed association gate to filter each object group, the fixed association gate may be too small, potentially filtering out first and second target objects that could be associated. Therefore, in this case, this embodiment can increase the baseline association gate to avoid preventing the first and second target objects with large errors from being unable to be associated.

[0074] Specifically, the distance between the object group consisting of the first target object and the second target object and the autonomous vehicle can be determined based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor.

[0075] In another possible implementation (Method 2), for any group of objects, the distance between the group and the autonomous vehicle is relatively close (less than a preset reference distance range), and the impact of distance factors is relatively small. In this case, the first target object (object feature information of the first target object) and the second target object (object feature information of the second target object) detected by the first and second sensors are relatively accurate. However, if the deviation of the fixed association gate compared to the object group is large, other interfering targets and noise may be included in the fixed association gate, causing the association between the first target object and the second target object to fail. Therefore, in this case, this embodiment can reduce the reference association gate to avoid including interfering factors that would prevent the first target object and the second target object from being accurately associated.

[0076] In another possible implementation (method 3), for any group of objects, the distance between the group of objects and the autonomous vehicle falls within the reference distance range. The first target object (object feature information of the first target object) and the second target object (object feature information of the second target object) detected by the first sensor and the second sensor are less affected by distance. In this case, there is no need to adjust the reference association gate.

[0077] S103, based on the dynamic association gate, determine whether the first target object and the second target object belong to the same object.

[0078] Specifically, for each object group, the deviation result between the feature information of each object in the object group is calculated; in one possible implementation, if the deviation result falls into the dynamic association gate corresponding to the object group, then the first target object and the second target object in the object group are considered to be the same object; otherwise, the first target object and the second target object in the object group are considered not to be the same object.

[0079] In another possible implementation, if the deviation result falls within the dynamic association gate corresponding to the object group, the object group is considered to have been filtered and can be further compared to determine whether the first target object and the second target object in the object group are the same object.

[0080] The aforementioned target association method introduces a dynamic association gate corresponding to distance factors by dynamically adjusting the baseline association gate, combining the first distance between the first sensor and the first target object detected by the first sensor, and the second distance between the second sensor and the second target object detected by the second sensor. This dynamic association gate reduces the impact of distance factors on target detection and improves the accuracy of target association when the first and second sensors have significant detection errors due to distance factors in the environment. Furthermore, by setting corresponding dynamic association gates specifically for each potentially associated object group (including the first and second target objects), the granularity of target association is refined, further improving association accuracy.

[0081] It is understandable that, based on the above S101-S103, the distance between each object group and the autonomous vehicle needs to be calculated. Only after the distance calculation can the corresponding dynamic association gate be matched. If there are many object groups, the amount of data calculation is large. Therefore, in this embodiment, corresponding filtering conditions are used to filter all objects in the first sensor and all objects in the second sensor, and the first target object and the second target object obtained are respectively used for error compensation only for the first target object and the second target object.

[0082] Specifically, the filtering criteria can be: for any object, the object has not been successfully associated with any other object during the filtering process based on a fixed association gate; correspondingly, the first target object and the second target object are the objects that have not been successfully associated with other objects in the above method 1 and the above method 2.

[0083] like Figure 2 As shown, this embodiment provides an optional method for extracting a first target object from first sensing data collected by a first sensor and a second target object from second sensing data collected by a second sensor, that is, a method for refining S101. The specific implementation process may include:

[0084] S201, extract a first candidate object from the first sensing data, and extract a second candidate object from the second sensing data.

[0085] In this embodiment, depending on the type of the first sensor and the second sensor, and the feature extraction algorithm corresponding to the first sensor and the second sensor, the first candidate object (or the second candidate object) can be an occluded object, an unoccluded object, or an object obtained by further analysis of the occluded object.

[0086] S202, based on the fixed association gate, the first candidate object and the second candidate object are filtered to obtain the first target object and the second target object.

[0087] Among them, the fixed associated gate can be the same as the reference associated gate, or it can be determined by the reference associated gate.

[0088] Specifically, the fixed association gate can effectively filter first and second candidate objects with relatively small errors, but may have association errors for first and second candidate objects with larger errors. Therefore, objects that cannot be successfully associated with other candidate objects are designated as first and second target objects, forming a first set C1 based on each first target object and a second set C2 based on each second target object. Then, each target object in the first set C1 and the second set C2 is further filtered using the corresponding dynamic association gates (generated in S101 to S103).

[0089] In this embodiment, during the screening process of fixed association gates, the first target object and the second target object that failed to be associated (the first candidate object and the second candidate object with larger errors) are obtained. The corresponding dynamic association gates are generated only for the corresponding first target object and the second target object, which reduces the amount of computation.

[0090] like Figure 3 As shown, this embodiment provides an optional method for adjusting the baseline association gate based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, to obtain a dynamic association gate, i.e., a method for refining S102. The specific implementation process may include:

[0091] S301, select at least one from the position error correlation gate, the size error correlation gate and the speed error correlation gate as the reference correlation gate.

[0092] Among them, the position error correlation gate can include the horizontal axis error correlation gate and the vertical axis error correlation gate; the size error correlation gate can include the length error correlation gate and the width error correlation gate; the speed error correlation gate can include the horizontal axis speed error correlation gate and the vertical axis speed error correlation gate.

[0093] Optionally, in this embodiment, one or more of the position error correlation gate, size error correlation gate, and velocity error correlation gate can be used as the reference correlation gate.

[0094] S302, determine the observation distance based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor.

[0095] Specifically, the sum or mean of the first distance and the second distance can be calculated, or a weighted distance value can be calculated by assigning different weights to the first distance and the second distance, and the calculated value can be used as the observation distance. Furthermore, for any object group, the observation distance determined based on the first target object and the second target object within the object group can be used as the distance between the object group and the autonomous vehicle.

[0096] S303, adjust the benchmark correlation gate according to the observation distance and the benchmark distance corresponding to the benchmark correlation gate to obtain the dynamic correlation gate.

[0097] Specifically, if the observation distance is greater than the reference distance (equivalent to the aforementioned reference distance range), the reference correlation gate is increased to obtain the dynamic correlation gate; if the observation distance is less than the reference distance, the reference correlation gate is decreased to obtain the dynamic correlation gate; the adjustment range of increasing or decreasing the reference correlation gate can be determined based on the difference or ratio between the observation distance and the reference distance.

[0098] Optionally, the ratio between the observed distance and the baseline distance can be used as an adjustment factor K; based on the adjustment factor, the baseline correlation gate can be scaled to obtain the dynamic correlation gate.

[0099] Specifically, the dynamic correlation gate is calculated as adjustment factor K × baseline correlation gate to scale the baseline correlation gate; adjustment factor K = observation distance / baseline distance; the above adjustment range is obtained based on the adjustment factor, or by further calculation of the adjustment factor.

[0100] In this embodiment, based on the proportional relationship between the observation distance and the reference distance, the reference correlation gate is magnified or reduced by a positive proportional coefficient (adjustment factor K) to obtain a more accurate dynamic correlation gate.

[0101] Furthermore, in this embodiment, since the baseline correlation threshold may include error thresholds in multiple dimensions, the resulting dynamic correlation gate also includes error thresholds in multiple dimensions. Therefore, when determining whether the deviation of any object group falls within the dynamic correlation gate, it can be determined whether the deviation of the object group falls within a set number of dynamic correlation gates. If so, the object group has passed the screening of the dynamic correlation gate; otherwise, the object group has not passed the screening of the dynamic correlation gate.

[0102] For example, the deviations of the object group include positional deviation, size deviation, and speed deviation. It is determined whether the number of deviations falling into the corresponding dynamic association gate meets the corresponding quantity. If it does, the object group passes the screening of the dynamic association gate.

[0103] like Figure 4 As shown, this embodiment provides an optional method for determining whether the first target and the second target objects belong to the same object pair based on a dynamic association gate, that is, a method for refining S103. The specific implementation process may include:

[0104] S401, based on dynamic association gate, determines the target association degree between the first target object and the second target object.

[0105] Among them, target relevance refers to the similarity between the first target object and the second target object.

[0106] Specifically, for any object group (including the first target object and the second target object), if the object group has passed the initial screening by the dynamic association gate (i.e., the deviation falls within the dynamic association gate), the target association degree of the object group is calculated. When calculating the target association degree, probabilistic similarity method, distance metric method, association coefficient method, etc. can be used. In this embodiment, the Mahalanobis distance between the first target object and the second target object is calculated, and the Mahalanobis distance is used as the target association degree. The Mahalanobis distance can include the Mahalanobis distance of the position information, the Mahalanobis distance of the velocity information, and the Mahalanobis distance of the size information between the target objects.

[0107] If the object group has not undergone preliminary screening by the dynamic association gate (i.e., the deviation falls within the dynamic association gate), then there is no need to calculate the similarity of the object group. The fixed value P can be directly used as the similarity result. Specifically, the similarity results of each target are shown in Table 2 below:

[0108] Table 2

[0109]

[0110] S402, determine whether the first target object and the second target object belong to the same object based on whether the target correlation degree meets the preset correlation conditions.

[0111] The preset association conditions can be either a similarity less than a preset value or a comparison result determined based on the target association degree of each object group.

[0112] Based on the comparison results determined by comparing each object group, to determine whether the association conditions are met, for ease of explanation, the following explanation uses any object group as an example. Specifically, as follows... Figure 5 As shown, this embodiment provides a refined process for determining whether the target correlation degree meets the preset correlation conditions, which may specifically include the following process:

[0113] S501, determine the first reference correlation degree between the first target object and the second reference object.

[0114] The second reference object is any object extracted from the second sensing data other than the second target object. Furthermore, for any object group, any second target object in the second set C2 other than the second target object within that object group can be used as the second reference object.

[0115] Specifically, the calculation method of the first reference correlation degree is the same as the calculation method of the target correlation degree. For example, when the first target object (1) and the second target object (1) are selected to form an object group, the similarity between the object groups is the target correlation degree. The first target object (1) and other second target objects, such as the second target object (2), the second target object (3), and the second target object (4), form each object group corresponding to the first reference correlation degree.

[0116] S502, determine the second reference correlation degree between the second target object and the first reference object.

[0117] Correspondingly, the first reference object is any object extracted from the first sensing data other than the first target object. Further, other first target objects in the first set C1, excluding the first target object within that object group, are used as the first reference objects.

[0118] Specifically, the calculation method of the first reference correlation degree is the same as the calculation method of the target correlation degree. For example, when the first target object (1) and the second target object (1) are selected to form an object group, the similarity between the object groups is the target correlation degree. The second target object (1) and other first target objects, such as the first target object (2), the first target object (3), and the first target object (4), form each object group corresponding to the second reference correlation degree.

[0119] S503, if the target correlation degree is less than both the first reference correlation degree and the second reference correlation degree, then the target correlation degree is determined to meet the preset correlation condition.

[0120] Specifically, the target correlation degree is compared with each first reference similarity and each second reference similarity. If the target correlation degree is the minimum value of its row and column in the correlation matrix corresponding to Table 2, then the target correlation degree is determined to meet the preset correlation conditions. Conversely, if the target correlation degree is not the minimum value of its row and column, then the target correlation degree is determined not to meet the preset correlation conditions.

[0121] Furthermore, if the target relevance does not meet the preset relevance conditions, then it is determined that the first target object and the second target object do not belong to the same object.

[0122] Furthermore, for a group of objects that meet the preset association conditions, in one possible approach, it can be directly determined that the first target object and the second target object in the group belong to the same object.

[0123] Since each candidate object in the first and second sensing data may be an occluded object, the first target object and the second target object may also be occluded objects. Therefore, in the case of large sensor error, in order to ensure the accuracy of the acquired associated targets, in another possible implementation, if the target association degree meets the preset association condition, then it is verified whether there is an occluded object in the object group that meets the association condition. If there is, the object group is filtered out as interference objects. If there is no, it is determined that the first target object and the second target object belong to the same object.

[0124] like Figure 6 As shown, this embodiment provides an optional method for determining whether a first target object and a second target object belong to the same object based on whether the target correlation satisfies a preset correlation condition, that is, a method for refining S502. The specific implementation process may include:

[0125] S601, Based on the recognition bounding box of the first target object and the recognition bounding box of the second target object, determine the error region.

[0126] The bounding box of the first target object is determined by the position and size information of the first target object, and the bounding box of the second target object is determined by the position and size information of the second target object.

[0127] Optionally, the identification box of the first target object (or the identification box of the second target object) can be a two-dimensional identification box or a three-dimensional identification box. In this embodiment, the identification box is a two-dimensional identification box, including the horizontal axis dimension, the horizontal axis position, the vertical axis dimension, and the vertical axis position.

[0128] Specifically, such as Figure 7As shown, the closed region formed by the preset position point on the autonomous vehicle (e.g., the center point of the rear axle), the recognition box of the first target object, and the recognition box of the second target object is used as the error region. The closed region includes at least the recognition box of the first target object and the recognition box of the second target object. Specifically, the edge of the error region can be determined by each point of each recognition box (or the line connecting the points).

[0129] S602, if there are no other objects besides the first target object and the second target object within the error region, then it is determined that the first target object and the second target object belong to the same object.

[0130] Among them, the other objects besides the first target object and the second target object can be any one of the first candidate object and the second candidate object.

[0131] Specifically, if other objects exist within the error area, it indicates that the first target object and / or the second target object in the object group may be occluded. In this case, the object group is treated as a false alarm and is not associated. If no other objects exist within the error area, it is determined that the first target object and the second target object in the object group belong to the same object.

[0132] For example, based on the above embodiments, this embodiment provides an optional example of an item recommendation method. For instance... Figure 8 As shown, the specific implementation process includes:

[0133] S801, extract a first candidate object from the first sensing data, and extract a second candidate object from the second sensing data;

[0134] S802, based on the fixed association gate, the first candidate object and the second candidate object are filtered to obtain the first target object and the second target object;

[0135] S803, select at least one of the position error correlation gate, the size error correlation gate and the velocity error correlation gate as the reference correlation gate;

[0136] S804, determine the observation distance based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor;

[0137] S805 uses the ratio between the observed distance and the reference distance as an adjustment factor;

[0138] S806, based on the adjustment factor, scales the baseline correlation gate to obtain the dynamic correlation gate;

[0139] S807, based on dynamic association gate, determines the target association degree between the first target object and the second target object;

[0140] S808, if the target correlation degree meets the preset correlation conditions, then the error region is determined based on the recognition box of the first target object and the recognition box of the second target object;

[0141] S809, if there are no other objects besides the first target object and the second target object within the error region, then it is determined that the first target object and the second target object belong to the same object.

[0142] Specifically, the process involves determining a first reference correlation degree between a first target object and a second reference object, wherein the second reference object is an object extracted from the second sensing data other than the second target object; determining a second reference correlation degree between the second target object and the first reference object, wherein the first reference object is an object extracted from the first sensing data other than the first target object; and determining that the target correlation degree satisfies a preset correlation condition if both the target correlation degree and the second reference correlation degree are less than the first and second reference correlation degrees.

[0143] The specific processes of S801-S809 described above can be found in the description of the above method embodiments. Their implementation principles and technical effects are similar, and will not be repeated here.

[0144] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0145] Based on the same inventive concept, this application also provides a target association apparatus for implementing the target association method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more target association apparatus embodiments provided below can be found in the limitations of the target association method described above, and will not be repeated here.

[0146] In one embodiment, such as Figure 9 As shown, a target association device 1 is provided, including an acquisition module 11, an adjustment module 12, and an association module 13, wherein:

[0147] The acquisition module 11 is used to extract a first target object from the first sensing data collected by the first sensor and to extract a second target object from the second sensing data collected by the second sensor.

[0148] The adjustment module 12 is used to adjust the reference association gate according to the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, to obtain the dynamic association gate;

[0149] The association module 13 is used to determine whether the first target object and the second target object belong to the same object based on the dynamic association gate.

[0150] In one embodiment, the adjustment module 12 includes:

[0151] The reference submodule is used to select at least one of the position error correlation gate, the size error correlation gate, and the velocity error correlation gate as the reference correlation gate;

[0152] The measurement submodule is used to determine the observation distance based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor;

[0153] The adjustment submodule is used to adjust the baseline association gate based on the observation distance and the baseline distance corresponding to the baseline association gate, so as to obtain the dynamic association gate.

[0154] In one embodiment, the adjustment submodule is further configured to: use the ratio between the observed distance and the reference distance as an adjustment factor;

[0155] Based on the adjustment factor, the baseline correlation gate is scaled to obtain the dynamic correlation gate.

[0156] In one embodiment, the associated module 13 includes:

[0157] The filtering submodule is used to determine the target correlation degree between the first target object and the second target object based on the dynamic association gate;

[0158] The matching submodule is used to determine whether the first target object and the second target object belong to the same object based on whether the target relevance meets the preset relevance conditions.

[0159] In one embodiment, the matching submodule is further configured to: if the target correlation degree meets the preset correlation conditions, determine the error region based on the identification box of the first target object and the identification box of the second target object;

[0160] If there are no other objects besides the first target object and the second target object within the error region, then the first target object and the second target object are determined to be the same object.

[0161] In one embodiment, the target association device further includes a comparison module, which is used to:

[0162] Determine a first reference correlation degree between the first target object and the second reference object; wherein, the second reference object is an object other than the second target object extracted from the second sensing data;

[0163] Determine a second reference correlation degree between the second target object and the first reference object; wherein, the first reference object is an object other than the first target object extracted from the first sensing data;

[0164] If the target correlation degree is less than both the first reference correlation degree and the second reference correlation degree, then the target correlation degree is determined to meet the preset correlation condition.

[0165] In one embodiment, the acquisition module 11 is further configured to: extract a first candidate object from the first sensing data, and extract a second candidate object from the second sensing data;

[0166] Based on the fixed association gate, the first candidate object and the second candidate object are filtered to obtain the first target object and the second target object.

[0167] Each module in the aforementioned target association device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0168] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a target association method. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0169] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0170] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0171] A first target object is extracted from the first sensing data collected by the first sensor, and a second target object is extracted from the second sensing data collected by the second sensor; a reference association gate is adjusted according to the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, to obtain a dynamic association gate;

[0172] Based on dynamic association gates, it is determined whether the first target object and the second target object belong to the same object.

[0173] In one embodiment, when the processor executes a computer program to adjust a reference correlation gate based on a first distance between a first target object and a first sensor, and a second distance between a second target object and a second sensor, to obtain a dynamic correlation gate, the following steps are specifically implemented: selecting at least one from a position error correlation gate, a size error correlation gate, and a velocity error correlation gate as a reference correlation gate; determining an observation distance based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor; and adjusting the reference correlation gate based on the observation distance and the reference distance corresponding to the reference correlation gate to obtain a dynamic correlation gate.

[0174] In one embodiment, when the processor executes a computer program to adjust the baseline correlation gate based on the observation distance and the baseline distance corresponding to the baseline correlation gate to obtain the logic of the dynamic correlation gate, the specific steps are as follows: the ratio between the observation distance and the baseline distance is used as an adjustment factor; based on the adjustment factor, the baseline correlation gate is scaled to obtain the dynamic correlation gate.

[0175] In one embodiment, when the processor executes a computer program to determine whether a first target and a second target object belong to the same object based on a dynamic association gate, the following steps are specifically implemented: determining the target association degree between the first target object and the second target object based on the dynamic association gate; and determining whether the first target object and the second target object belong to the same object based on whether the target association degree meets a preset association condition.

[0176] In one embodiment, when the processor executes the logic of whether the target correlation degree of the computer program meets the preset correlation conditions and determines whether the first target object and the second target object belong to the same object, the following steps are specifically implemented: if the target correlation degree meets the preset correlation conditions, then an error region is determined based on the identification box of the first target object and the identification box of the second target object; if there are no other objects besides the first target object and the second target object in the error region, then it is determined that the first target object and the second target object belong to the same object.

[0177] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining a first reference correlation degree between a first target object and a second reference object; wherein the second reference object is an object other than the second target object extracted from the second sensing data; determining a second reference correlation degree between the second target object and the first reference object; wherein the first reference object is an object other than the first target object extracted from the first sensing data; if both the target correlation degree and the second reference correlation degree are less than the first reference correlation degree and the second reference correlation degree, then determining that the target correlation degree satisfies a preset correlation condition.

[0178] In one embodiment, when the processor executes the logic of a computer program to extract a first target object from first sensing data collected by a first sensor and to extract a second target object from second sensing data collected by a second sensor, the processor specifically implements the following steps: extracting a first candidate object from the first sensing data and extracting a second candidate object from the second sensing data; filtering the first candidate object and the second candidate object based on a fixed association gate to obtain the first target object and the second target object.

[0179] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0180] A first target object is extracted from the first sensing data collected by the first sensor, and a second target object is extracted from the second sensing data collected by the second sensor; a reference association gate is adjusted according to the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, to obtain a dynamic association gate;

[0181] Based on dynamic association gates, it is determined whether the first target object and the second target object belong to the same object.

[0182] In one embodiment, when the computer program adjusts the baseline association gate based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, and the logic of the dynamic association gate is executed by the processor, the following steps are specifically implemented:

[0183] At least one of the position error correlation gate, size error correlation gate, and velocity error correlation gate is selected as the reference correlation gate; the observation distance is determined based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor; the reference correlation gate is adjusted based on the observation distance and the reference distance corresponding to the reference correlation gate to obtain the dynamic correlation gate.

[0184] In one embodiment, when the logic of adjusting the baseline correlation gate based on the observation distance and the baseline distance corresponding to the baseline correlation gate to obtain the dynamic correlation gate is executed by the processor, the following steps are specifically implemented: using the ratio between the observation distance and the baseline distance as an adjustment factor; scaling the baseline correlation gate based on the adjustment factor to obtain the dynamic correlation gate.

[0185] In one embodiment, when the logic of determining whether a first target and a second target object belong to the same object based on a dynamic association gate is executed by the processor, the following steps are specifically implemented: determining the target association degree between the first target object and the second target object based on the dynamic association gate; determining whether the first target object and the second target object belong to the same object based on whether the target association degree meets a preset association condition.

[0186] In one embodiment, when the logic of a computer program determining whether a first target object and a second target object belong to the same object based on whether the target correlation degree meets a preset correlation condition is executed by the processor, the following steps are specifically implemented: if the target correlation degree meets the preset correlation condition, then an error region is determined based on the identification box of the first target object and the identification box of the second target object; if there are no other objects besides the first target object and the second target object in the error region, then it is determined that the first target object and the second target object belong to the same object.

[0187] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining a first reference correlation degree between a first target object and a second reference object; wherein the second reference object is an object other than the second target object extracted from the second sensing data; determining a second reference correlation degree between the second target object and the first reference object; wherein the first reference object is an object other than the first target object extracted from the first sensing data; if both the target correlation degree and the second reference correlation degree are less than the first reference correlation degree and the second reference correlation degree, then determining that the target correlation degree satisfies a preset correlation condition.

[0188] In one embodiment, when the logic of extracting a first target object from first sensing data collected by a first sensor and extracting a second target object from second sensing data collected by a second sensor is executed by a processor, the specific steps are as follows: extracting a first candidate object from the first sensing data and extracting a second candidate object from the second sensing data; filtering the first candidate object and the second candidate object based on a fixed association gate to obtain the first target object and the second target object.

[0189] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0190] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0191] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0192] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A target association method, characterized in that, The method includes: Extract a first target object from the first sensing data collected by the first sensor, and extract a second target object from the second sensing data collected by the second sensor; Based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, the baseline correlation gate is adjusted to obtain the dynamic correlation gate; wherein, the baseline correlation gate is used to describe the maximum permissible deviation between the first target object and the second target object in a specified dimension; Determine the deviation between the object feature information of the first target object and the object feature information of the second target object; If the deviation result falls within the dynamic correlation gate, the target correlation degree between the first target object and the second target object is determined; Based on whether the target correlation degree meets the preset correlation conditions, it is determined whether the first target object and the second target object belong to the same object.

2. The method according to claim 1, characterized in that, The step of adjusting the baseline association gate based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor, to obtain the dynamic association gate includes: Select at least one of the position error correlation gate, the size error correlation gate, and the velocity error correlation gate as the reference correlation gate; The observation distance is determined based on the first distance between the first target object and the first sensor, and the second distance between the second target object and the second sensor; The reference correlation gate is adjusted based on the observation distance and the reference distance corresponding to the reference correlation gate to obtain the dynamic correlation gate.

3. The method according to claim 2, characterized in that, The step of adjusting the benchmark correlation gate based on the observation distance and the benchmark distance corresponding to the benchmark correlation gate to obtain the dynamic correlation gate includes: The ratio between the observed distance and the reference distance is used as an adjustment factor; Based on the adjustment factor, the baseline correlation gate is scaled to obtain the dynamic correlation gate.

4. The method according to claim 1, characterized in that, The step of determining whether the first target object and the second target object belong to the same object based on whether the target correlation degree meets the preset correlation conditions includes: If the target correlation degree meets the preset correlation conditions, then the error region is determined based on the recognition box of the first target object and the recognition box of the second target object; If there are no other objects besides the first target object and the second target object within the error region, then the first target object and the second target object are determined to be the same object.

5. The method according to claim 1, characterized in that, The step of determining whether the first target object and the second target object belong to the same object based on whether the target correlation degree meets the preset correlation conditions includes: If the target correlation degree meets the preset correlation conditions, then verify whether there is an occluded object among the first target object and the second target object; If neither the first target object nor the second target object contains an occluded object, then the first target object and the second target object are determined to be the same object.

6. The method according to claim 4, characterized in that, The method further includes: A first reference correlation degree is determined between the first target object and the second reference object; wherein, the second reference object is an object other than the second target object extracted from the second sensing data; Determine a second reference correlation degree between the second target object and the first reference object; wherein, the first reference object is an object other than the first target object extracted from the first sensing data; If the target correlation degree is less than both the first reference correlation degree and the second reference correlation degree, then the target correlation degree is determined to meet the preset correlation condition.

7. The method according to claim 1, characterized in that, Extracting the first target object from the first sensing data collected by the first sensor, and extracting the second target object from the second sensing data collected by the second sensor, includes: Extract a first candidate object from the first sensing data, and extract a second candidate object from the second sensing data; Based on the fixed association gate, the first candidate object and the second candidate object are filtered to obtain the first target object and the second target object.

8. A target association device, characterized in that, The device includes: The acquisition module is used to extract a first target object from the first sensing data collected by the first sensor, and to extract a second target object from the second sensing data collected by the second sensor. An adjustment module is used to adjust a baseline correlation gate based on a first distance between the first target object and the first sensor, and a second distance between the second target object and the second sensor, to obtain a dynamic correlation gate; wherein the baseline correlation gate is used to describe the maximum permissible deviation between the first target object and the second target object in a specified dimension; The association module is used to determine the deviation result between the object feature information of the first target object and the object feature information of the second target object; if the deviation result falls within the dynamic association gate, it determines the target association degree between the first target object and the second target object; and determines whether the first target object and the second target object belong to the same object based on whether the target association degree meets the preset association conditions.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images